Download A mixed-signal sensor interface microinstrument

Sensors and Actuators A 91 (2001) 266±277
A mixed-signal sensor interface microinstrument
Keith L. Kravera,*, Matthew R. Guthausa, Timothy D. Stronga, Peter L. Birda,
Geun Sig Chab, Wolfgang HoÈldc, Richard B. Browna
a
Department of Electrical Engineering and Computer Science, The University of Michigan, Ann Arbor, MI 48109, USA
b
Chemical Sensor Research Group, Department of Chemistry, Kwangwoon University, Seoul, South Korea
c
Information Appliance Division, National Semiconductor Corporation, D-82256 FuÈrstenfeldbruck, Germany
Abstract
A single-chip implementation of a microinstrumentation system is presented. The chip incorporates voltage, current, and capacitive
sensor interfaces; a temperature sensor; a 10-channel, 12-bit analog-to-digital converter; and an 8-bit microcontroller with a 16-bit hardware
multiplier and a 40-bit accumulator. Serial and parallel interfaces allow digital communication with a host system. Fabricated in a standard
0.35 mm digital CMOS process, the die occupies 3:8 mm 4:1 mm, operates from a nominal supply voltage of 3 V, and draws 16 mA when
fully powered (850 mA standby current). To facilitate testing of the prototype, extra pads are bonded out to package pins. The chips are
packaged in 132-pin ceramic pin-grid-array packages. # 2001 Elsevier Science B.V. All rights reserved.
Keywords: Sensor interfaces; Microinstrumentation; Mixed-signal design; Embedded microcontroller
1. Introduction
Sensor interfaces have been identi®ed as critical to the
development of the MEMS ®eld [1]. Multi-chip sensor/
interface-circuit solutions have been developed [2], and
interface electronics have been integrated with sensors
[3]. This paper presents a single-chip microinstrument: a
mixed-signal microcontroller with an analog-front-end
including voltage, current, and capacitive sensor interfaces;
signal ampli®cation; and analog-to-digital conversion circuitry. The only sensor integrated on this chip measures
temperature, but other sensors and actuators could be
included. However, for many applications in which the
MEMS devices cannot be made with CMOS process steps,
or in which sensor lifetime is limited, the most ef®cient
system partitioning allocates all or most of the electronics to
a microinstrument chip separate from the sensor.
Recent advances in mixed-signal chips that target sensor
applications have been reported. Cirrus Logic developed an
integrated circuit (IC) that combines an input multiplexor, a
programmable gain instrumentation ampli®er, a high resolution ADC, and a serial interface to communicate with an
external microcontroller [4]. The programmability of gain
settings and ®ltering cutoff frequencies facilitates a variety
of sensor interface circuits. Designed in a 0.6 mm CMOS
*
Corresponding author.
process, the chip requires a 5.0 V supply. Analog Devices
released a data acquisition (DAQ) IC that includes a temperature sensor, a pair of digital-to-analog converters, and an
8052 microcontroller core in addition to the blocks in the
Cirrus Logic chip [5] Ð a single-chip data acquisition
system. The Analog Devices IC, the MicroConverterTM,
supports multiple communication bus protocols, contains
¯ash EEPROM for data and program storage, and is speci®ed to operate with a supply voltage down to 3.0 V. Another
reported single-chip DAQ system includes support for the
IEEE 1451 standard transducer interface in addition to the
basic architecture of the Analog Devices IC [6]. This chip,
designed in a 0.6 mm CMOS process (with embedded
EEPROM), operates down to a supply voltage of 2.7 V
and facilitates in-circuit emulation. This paper describes a
mixed-signal microcontroller which was designed speci®cally for use with solid-state sensors. The presented microinstrument increases the integration level over the ICs
reported above by combining multiple sensor interface
circuits, the data acquisition system, and the microcontroller
on a single die.
This mixed-signal microinstrument, named the MS-8, is
optimized, from instruction set de®nition to analog functionality, to economically support embedded sensor applications. The MS-8 supports these applications by increasing
integration at the chip level, supporting in-circuit emulation
techniques, employing system power reduction methodologies, and enhancing data acquisition and signal processing
0924-4247/01/$ ± see front matter # 2001 Elsevier Science B.V. All rights reserved.
PII: S 0 9 2 4 - 4 2 4 7 ( 0 1 ) 0 0 5 9 6 - 9
K.L. Kraver et al. / Sensors and Actuators A 91 (2001) 266±277
267
Fig. 1. MS-8 block diagram. The analog (shaded) and digital (unshaded) sections communicate through the analog interface block, treated as a peripheral by
the processor core.
capabilities beyond simple microcontrollers. Also, a minimal version of the microinstrument can be implemented
with as few as eight pins. See the block diagram in Fig. 1 [7].
To allow in-circuit emulation, the MS-8 includes hardware support for a single breakpoint or trace event; a singlepin development system interface; and instructions for
halting the processor, single stepping through instructions,
and reading/writing system registers. Single-cycle interrupt
response and direct memory access (DMA) capabilities
facilitate usage of the MS-8 in time-critical applications,
and a 16-bit 16-bit hardware multiply, 40-bit accumulate
block provides signi®cant signal processing capability for a
small microcontroller. To minimize power dissipation, data
paths were limited to 8-bit widths, and the chip was designed
to operate with a 3.0 V supply, in a 0.35 mm CMOS process.
In addition, a clock manager provides programmability of
the clock frequency for the digital and analog circuits, and
each analog block can be individually powered down under
software control. This paper describes the digital and analog
functionality of the MS-8 and presents sensor measurements
taken with the microinstrument.
2. Digital core and peripherals
The right side of Fig. 1 depicts the digital portion of the
MS-8, which includes the processor core, peripherals, and
memory [8]. The instruction set architecture (ISA) of the
MS-8 is a blend of reduced instruction set computer (RISC)
and complex instruction set computer (CISC) architectures.
Instruction lengths vary in length, ranging from one to four
bytes to increase code density, but the architecture is load/
store to decrease control and data path complexity. The
instruction encoding assigns the most frequently used
instructions fewer bytes, decreasing the size of the program.
Another bene®t of the load/store architecture is that compilers can ef®ciently implement ANSI C. The simpli®ed
control and data path logic conserves power by not implementing unused or infrequently used addressing modes.
This, coupled with increased code density decreases the
overall power.
Program memory in this prototype version of the MS-8
consists of a 512 byte boot ROM and a 4KB RAM; a
512 byte RAM provides data storage. For application and
testing versatility, both RAMs can be used as either instruction or data storage. In addition, off-chip memories can
occupy sections of the 20-bit address space. Complete
support for a single external memory segment is included
on-chip; no board-level logic is required.
The MS-8 has a standard bus speci®cation that allows
third party soft IP peripherals to be easily integrated. The
MS-8 interrupt arbitration is done with a distributed scheme
using wired-OR signaling over the address and data bus.
This reduces the number of signals between modules by
268
K.L. Kraver et al. / Sensors and Actuators A 91 (2001) 266±277
reusing existing wires and distributes the priority arbitration
logic among all peripherals. Therefore, a system with only a
few peripherals will not have the entire priority arbitration
logic wasting area in the core.
To reduce dynamic power consumption, the MS-8 has a
programmable clock manager (CLKMAN) that can divide
the global clock frequency by up to 16, select a slow external
clock, or put the system to sleep until a certain event occurs.
While in the sleep mode, the global clock is stopped to all
units except the timer/watchdog manager and the analog
section. The clock manager also generates a four-phase nonoverlapping clock for the DS ADC and other switchedcapacitor circuits.
The parallel input/output (PIO) peripheral reads/writes 8bit parallel data using the DMA functionality of the core.
The PIO is programmed with an address and a number of
bytes. After con®guration, the PIO transfers or receives the
data, signaling an interrupt at the end of the ®nal byte.
The multiply accumulate unit (MAC) is capable of performing a single cycle multiply and accumulate. The two
operands are 16-bit registers, and the accumulator is a 40-bit
register. Without the MAC, a 16-bit 16-bit multiply accumulate would take many cycles to complete using a Booth
(or similar) multiply scheme. In the MS-8 architecture, the
Booth scheme would require multi-byte addition instructions, because the data path is only 8 bits. However, using
the MAC, with its parallel multiplication, a multiply and an
accumulate complete in a single cycle.
The analog interface block (AIB) peripheral includes 11
control registers and ®ve data registers for the analog
circuits. Control registers store con®guration settings
including which units to turn off, gain settings, and analog-to-digital (ADC) converter resolution. Data registers
buffer samples from the ADCs and can be con®gured to
generate interrupts after each conversion.
The universal synchronous/asynchronous receiver/transmitter (USART) performs synchronous or asynchronous
serial communication. Programmable baud rates range from
4800 bps to 2.5 Mbps at a system frequency of 40 MHz. The
serial protocol uses a single start bit and a programmable
stop bit of 1, 1.5 or 2 bits. The USART also has programmable support for odd, even, or no parity.
The multifunction timer (MFT) employs two independent
16-bit counters, which can be used together for pulse-width
modulation or as independent timers, or for input event
capture. The clock sources include the pre-scaled system
clock, the slow system clock, or a gated external clock for
determining pulse widths. Each timer has a programmable
interrupt level.
The timer/watchdog manager (TWM) is similar to the
MFT except that the latter only has a single 16-bit counter.
This timer operates on a pre-scaled system clock and is not
stopped during sleep mode. The timer is used to generate
wake-up events when the system is put to sleep and to
recover from system faults. The TWM can also use synchronized external events as wake-up signals.
3. Analog core
The analog front-end (AFE), outlined in the left half of
Fig. 1, includes a multiplexor which selects from an array of
buffered voltage inputs (for potentiometric chemical sensors), the temperature sensor, a capacitive sensor interface
(for pressure and acceleration sensors), and an amperometric
chemical sensor interface. Signal conditioning provided by
the programmable gain ampli®er (PGA) makes the interface
generic and increases the dynamic range of the system. A
passive low-pass ®lter and a 12-bit analog-to-digital converter (ADC) complete the AFE. The analog functionality is
further described in the remainder of this section.
The overriding goal for this prototype was to keep the
design simple. This concept is re¯ected in the choice of
architectures for the interface and signal processing circuits
and the pervasive use of a general purpose opamp. For
example, the analog core employs single-ended circuits,
which have the bene®t of power reduction, but are more
susceptible to digital circuit induced substrate noise. Keeping the design simple reduced the design time and provided a
baseline of performance upon which future versions can
improve.
3.1. Temperature sensor
As shown in Fig. 2, the temperature sensor is comprised of
a proportional to absolute temperature (PTAT) current
source, represented by the circle, and a transimpedance
ampli®er. The PTAT current source, shown in Fig. 3, generates a current that is proportional to the thermal voltage,
kT/q. M1±M4, Q1±Q2, and R1 produce a current according
to
I1 …T† ˆ
1 kT
ln …A†;
R1 q
where A is the emitter area ratio of Q2 to Q1. Transistors M5,
M6, and Q3 provide feedback to reduce the effects of the
finite output resistance of M4; Cc compensates the feedback
Fig. 2. Simplified schematic of the temperature sensor. The temperature
sensor outputs a voltage derived from a current proportional to the thermal
voltage.
K.L. Kraver et al. / Sensors and Actuators A 91 (2001) 266±277
269
the interface circuitry provides actuation to the sensor in
addition to readout. The remainder of this section describes
the interface circuitry included in this implementation of the
MS-8.
3.2.1. Voltage buffers
The buffered voltage inputs provide a high-impedance
interface for potentiometric sensors (and other voltage
signals) and accommodate rail-to-rail signals (0±3.0 V).
Potentiometric chemical sensors have output impedances
on the order of 100 MO; therefore, they require readout
circuitry with high input impedance. A CMOS rail-to-rail
opamp con®gured as a voltage follower provides the necessary high input impedance buffer. The opamp is described
later. Buffers not in use can be powered down selectively.
Fig. 3. Current source used in the temperature sensor. The output current
is proportional to absolute temperature (kT/q).
loop to ensure stability. Ms1±Ms4 and Qs1 direct the circuit
to the correct operating point upon startup. Since
…W=L†M3 ˆ …W=L†M4 ˆ …W=L†M6 ˆ 1=B…W=L†M7 ; I0 …T† ˆ
BI1 …T†.
The temperature sensor in Fig. 2 compares the PTAT
current to a reference current and generates a voltage
according to
V…T† ˆ
B
Rf
kT
Rf
ln…A† ‡ Vref 1 ‡
;
q
R1
Rref
where Vref is a temperature-independent voltage generated
by the on-chip bandgap reference. The values of the constants and resistors are chosen to set the output voltage in the
middle of the supply range for room temperature and to set
the gain. Note that since all resistors are constructed from
p‡ diffusion and appear in ratios in the above equation,
temperature coefficients cancel. Temperature range and
sensitivity are competing characteristics controlled by the
gain of the programmable gain amplifier, discussed in
Section 3.4.
This type of temperature sensor is employed in the MS-8
for two reasons. First, the PTAT current source also supplies
bias current to the other analog circuit blocks, making
double use of the circuit with its associated power dissipation and area. Second, the linear nature of the output voltage
response makes calibration straightforward.
3.2.2. Capacitive sensor readout
The capacitive interface circuit converts a capacitance to a
voltage. Capacitive-based pressure sensors and accelerometers experience a capacitance change in response to
an applied force [2]. In the readout circuit shown in Fig. 4
[9], a switched-capacitor charge integrator compares the
resulting sensor capacitance to a reference capacitor to
produce a voltage as follows:
Vout ˆ VA
CS
Cref
Cf
‡ Vref ;
where CS and Cref are the sensor and reference capacitors,
respectively; VA is the amplitude of the clock, f; and Vref is
the on-chip reference voltage. This architecture was chosen
because of the circuit's insensitivity to input parasitic capacitances and ease of implementation. Note that the circuit
provides ports for up to four capacitive sensors. As the
equation shows, the readout circuit generates a voltage
linearly proportional to the sensor capacitance.
When f transitions to the high state, the reset switch on
the integrator is closed and the opamp charges the sensor
capacitance. When f falls, the feedback capacitor, Cf integrates a charge proportional to the difference between the
sensor and reference capacitors. The MS-8 provides several
3.2. Sensor interfaces
The primary function of sensor interface circuitry is to
read and amplify the transducer output signals. As described
below for capacitive sensors, the readout operation can
involve the conversion of a capacitance signal to an electrical signal (voltage). Some sensors require, or their performance can be improved by, closed loop operation, where
Fig. 4. Simplified circuit diagram of the capacitive readout circuit. The
circuit facilitates recording of up to four external capacitive sensors and
has a programmable reference capacitor bank.
270
K.L. Kraver et al. / Sensors and Actuators A 91 (2001) 266±277
Fig. 5. Simplified diagram of the amperometric sensor interface circuit and off-chip sensor. The circuit applies a triangular wave to the sensor to facilitate
cyclic voltammetry, or can use any externally supplied voltage to drive the amperometric cell.
options for the reference capacitor: an on-chip programmable capacitor array, an external capacitor, or a combination of both. The internal capacitor array has capacitors in
the range of 250 f F to 15 pF in the steps of 250 fF. This
granularity is insuf®cient for general applications, but is
suitable for demonstration of the prototype: the reference
capacitor should be close in value to the sensor capacitance
so as to not saturate the output voltage.
3.2.3. Amperometric sensor interface
Amperometric sensors are frequently employed to detect
dissolved gases or uncharged molecules in liquid. As shown
in Fig. 5, these sensors typically include three electrodes:
counter, reference, and working [10]. The sub-circuit containing opamps 1 and 2 and resistors R1 and R2 form a
potentiostat circuit which employs feedback to regulate the
voltage applied across the reference and working electrodes,
independent of the current through the cell. In this version of
the MS-8, a simple triangular wave generator, consisting of
opamp 3 and the external capacitor, Cext, generates an
applied voltage suitable for cyclic voltammetry. For other
types of amperometry, the user can stop the clock under
program control and provide an external voltage in place of
the capacitor to excite the sensor with arbitrary waveforms.
For example, Kissinger and Heineman [11] describe many
types of amperometry, including differential pulse and
square-wave. The next generation MS-8 will include a
digital-to-analog converter to generate arbitrary waveforms
without the use of an off-chip capacitor. Current sensing
occurs at the working electrode with a simple transimpedance ampli®er which generates a voltage proportional to
the cell current (opamp 4). To accommodate sensors having
differing current magnitudes, Rf can be programmed through
the AIB or supplied off chip. The prototype has only two
resistor settings, 3.75 and 242 kO, suitable for currents in the
range of 400 mA to 100 nA (with the use of the PGA). More
on-chip resistor values are needed to facilitate a wider range
of sensors Ð most of the measurements reported herein
were with an external resistor.
3.3. Analog multiplexor
The 10:1 differential analog mux consists of two 10:1
single-ended muxes. Each of the previously mentioned mux
inputs feeds a channel of both single-ended muxes. With this
connection scheme, the mux provides maximum versatility
in selecting inputs. The AIB contains a register and logic for
con®guring the mux under program control.
Measurements appear to indicate crosstalk (capacitive
coupling) between channels. To combat this problem, the
next generation MS-8 will ground the mux channels not
selected.
3.4. Programmable gain amplifier
The programmable gain ampli®er ideally boosts the input
signal to the full-scale range of the analog-to-digital converter. In this implementation of the MS-8, the PGA is
designed after the standard three-opamp, two-stage instrumentation ampli®er (see Fig. 6). The ®rst stage provides
Fig. 6. Programmable instrumentation amplifier. Gain is set by an on-chip
programmable resistor (R1).
K.L. Kraver et al. / Sensors and Actuators A 91 (2001) 266±277
high input impedance, common-mode rejection, a differential input, and programmable gains of 1±61 in steps of 10 V/
V. Testing with actual sensors has revealed that smaller gain
steps are needed to realize the full dynamic range of the
system. Stage two of the ampli®er provides low output
impedance, differential to signal-ended conversion, and
level shifting. Use of the low noise, rail-to-rail opamp
(described later) in both stages increases the dynamic range
of the PGA. At the output of the PGA, a simple low-pass
®lter reduces ampli®er distortion and aliasing (resulting
from the analog-to-digital converter sampling process). A
control register in the AIB stores the gain settings.
The drawbacks of the three-opamp architecture include
area and power penalties. These penalties are exacerbated by
the fact the opamp is designed for general purpose use and is
thus larger and consumes more power than required for this
application.
3.5. Bandgap reference
The bandgap reference, based on the kT/q current source
of Fig. 3, supplies bias current and a voltage reference for the
analog circuitry. An opamp in a negative feedback con®guration provides voltage gain to boost the reference voltage
to mid-rail and buffering to drive the reference to the DS
ADC, programmable gain ampli®er, temperature sensor,
potentiostat, and off chip. To avoid start-up transient behaviors, the bandgap circuit is powered even in low-power
mode; the buffer, however, has power-off support.
3.6. Analog-to-digital converters
Two selectable data converters produce a digital representation of the analog signal. The single-slope ADC has
higher resolution than the DS ADC but is more susceptible
to substrate noise. Because of this limitation, the singleslope converter will be useful for studying noise-reduction
methods in this mixed-signal environment. The ramp voltage for the converter is generated by supplying a ®xed
current, derived from the bandgap reference, to an off-chip
capacitor.
The ®rst-order DS ADC (the primary data converter) is
shown in Fig. 7. DS modulators trade bandwidth for resolution, which makes the architecture suitable for this
application, which requires moderate to high resolution data
conversion of low frequency sensor outputs. In addition, the
combination of integration and the feedback reduces in-band
substrate noise produced by the switching activity of the
digital circuits. The DS architecture also relaxes precision
requirements on the analog circuitry, which makes this type
of ADC suitable for design in a digital process [12].
A counter in the AIB ®lters and decimates the oversampled, single-bit ADC output. The default counter width
is 12 bits, but the user can con®gure the counter to any width
up to and including 16 bits. The AIB stores both the counter
output and the single-bit oversampled output, allowing the
271
Fig. 7. Simplified diagram of the first-order DS modulator. Filtering is
provided by the counter or with software and the MAC peripheral.
processor core to provide the ®ltering and decimation, if
desired by the user. The counter requires minimal area, but
forces an oversampling ratio of 2n for n-bit data, which is not
an ef®cient use of power and puts stringent demands on the
unity gain frequency and slew rate of the opamp in the
integrator. More sophisticated decimation schemes could be
implemented in the microcontroller core.
Under program control, the clock manager divides the
system clock to generate the sampling clock and permits the
user to vary the sampling frequency according to the application. In addition, the clock manager generates the four
clock phases shown in Fig. 7. This clocking scheme reduces
the impact of switch-induced charge injection at the high
impedance nodes.
3.7. Rail-to-rail opamp
The custom opamp of Fig. 8, based on a topology in [13],
was designed for use in the analog circuits described previously. Since in some cases the opamp drives the signal off
chip (e.g. the reference voltage and test points), the opamp is
designed to drive 200 pF loads. To maximize dynamic range
in this low-voltage environment, the opamp processes railto-rail input signals and drives rail-to-rail output signals.
Transistors M30±M32 maintain constant input transconductance, gmi over the input common mode range by maintaining the sum of the currents in the two input differential pairs
(M1±M4) constant. A constant gmi reduces distortion and
makes the Miller compensation more ef®cient in terms of
capacitor area and overall power dissipation. The transistors
in the input differential pairs and output stage (M26 and
M27) operate in weak inversion to reduce input referred
noise and offset and to maximize input- and output-stage
transconductance for a given current. M19±M24 provide
class AB biasing to the output stage to reduce power.
This opamp is designed to have low input referred offset
voltage and noise over the full input common mode range, a
temperature range of 40 to 858C, and a supply range of
2.5±3.3 V. Since the opamp drives off-chip signals, the
output stage consumes around 75% of the total power to
272
K.L. Kraver et al. / Sensors and Actuators A 91 (2001) 266±277
Fig. 8. Simplified schematic of the rail-to-rail opamp without bias circuits. The input differential pairs (M1±M4) and output stage (M25 and M26) operate in
weak inversion to reduce input referred noise, offset, and power.
provide suf®cient unity gain phase margin. As mentioned
before, this opamp is used throughout in the analog circuit
blocks, even where load capacitance is minimal. A more
ef®cient design, in terms of area and power, would be
realized by employing a custom ampli®er for each block.
The MS-8 does not include any offset control in the
readout channels, which could potentially lead to saturation
if high gains are employed. But, as mentioned above, the
opamp is designed to have a low offset and measurements
reveal that the offset drift with temperature is also low
relative to a 3.0 V supply (see Section 5.1). Future implementations of the MS-8 will likely provide offset cancellation, especially for lower voltage operation.
4. Physical design
Fig. 9 is a die photomicrograph of the MS-8. The major
components visible are the RAMs, ROM, analog circuitry,
and the microcontroller core. The die is approximately
3:8 mm 4:1 mm. Nearly one-quarter of the core area,
1:39 mm 1:42 mm, is dedicated to the analog circuits.
RAM and ROM memory cells occupy a little less than half
of the core. The die was fabricated in a 0.35 mm digital
CMOS process and contains over 300,000 transistors. For
testing purposes, 103 signal pins and 12 power pins were
bonded out. However, some applications would require as
few as eight pads to be bonded to pins.
The memory drivers are placed as far from the analog
circuits as possible to reduce substrate coupling effects. In
addition, the analog core has ample substrate contacts
surrounding the sensitive circuitry, and dedicated VDD and
ground connections to further combat switching noise.
Finally, enforcing greater-than-minimum metal-to-metal
spacings for sensitive signals reduces crosstalk between
interconnect lines (at the expense of area).
5. Experimental results
The MS-8 digital core was veri®ed to be functional at
40 MHz, 3.0 V, and room temperature using an HP82000
digital tester. While simulations predict static current draw
around 3.3 mA (full power) and 15 mA (standby), measurements reveal 16 mA and 850 mA currents, respectively.
Table 1 shows simulated power for each analog circuit block
and the total measured power. Note that most of the unexpected power dissipation is from the digital circuits. The
authors suspect leakage in the RAMs. The discrepancy
between the simulated and measured power in the analog
core is also under investigation.
Initial testing of the AFE demonstrates the usefulness of
the device in embedded sensor applications. The USART
supports bidirectional serial communications between the
MS-8 and a laptop computer running LabView (National
Instruments); programs were downloaded to the chip and
digital samples from the ADC were transferred to LabView
for recording and analysis over this serial link. The remaining sections report initial results of measurements of potentiometric and amperometric chemical sensors, pressure
sensors, and the temperature sensor. All the sensor interfaces
and digital components have been veri®ed to be functional.
The opamp has also been characterized; the results of which
are described ®rst.
5.1. Opamp characterization
Table 2 shows the typical opamp characteristics. These
measurement were taken under nominal conditions (room
temperature, 3.0 V supply) from opamps packaged separately. The open-loop gain, slew rate, and unity gain bandwidth are suf®cient to enable oversampling ratios on the
ADC above 212 and sampling rates in excess of 1 MHz,
according to [12].
K.L. Kraver et al. / Sensors and Actuators A 91 (2001) 266±277
273
Fig. 9. MS-8 die photomicrograph. The chip was fabricated in a 0.35 mm national semiconductor process and measures 3:8 mm 4:1 mm.
5.2. Potentiometric sensor results
Fig. 10 shows the calibration curve for over seven decades
of KCl concentration for a solid-state potentiometric sensor
selective to potassium ions (K‡) [14]. A commercial doublejunction reference electrode was used for the reference. The
inset contains the time response of the sensor; each step
represents a 10-fold increase in potassium concentration. An
external electrode connected to the on-chip reference generator sets the potential of the solution to 1.5 V to keep the
signal in the range of the interface circuitry. The sensor has a
linear response from around the minimum detectable limit of
10 5 M to about 1 M, as shown in the ®gure. Using the
microcontroller to linearize the response with a higher-order
®t would extend the detection range. Theory predicts a slope
of 59 mV per decade; non-idealities in the sensor and gain
error in the PGA are possible reasons for the slope discrepancy.
5.3. Amperometric sensor results
In a class of amperometry called chronoamperometry, a
voltage is applied between the working and reference electrodes of suf®cient potential to cause oxidation or reduction
Table 1
Measured and simulated power dissipation for the analog corea
Unit
Simulated
(mW)
Buffer
Temperature sensor
Capacitive sensor interface
Amperometric sensor interface
Bandgap (buffered)
PGA
DS ADC
540
570
550
2160
540
1620
736
All units on
All units off
9416
45
a
Measured
(mW)
Conditions
(fsample)
100 kHz
1 MHz
9154
420
Note that the bandgap reference circuit always remains powered.
Fig. 10. K‡ calibration curve. The background solution is 0.05 M trizma
base (Sigma, St. Louis, MO) adjusted to pH 7.4 with H2SO4. The PGA
gain is set to 1 V/V.
274
K.L. Kraver et al. / Sensors and Actuators A 91 (2001) 266±277
Table 2
Typical measured characteristics of the CMOS opampa
Symbol
Parameter
VOS
Offset voltage
Offset drift
Positive power supply rejection ratio
Negative power supply rejection ratio
Common mode rejection ratio
Open-loop gain
Slew rate
Gain-bandwidth
Supply current
Phase margin
Input referred
Voltage noise
PSRR‡
PSRR
CMRR
AO
SR
GBW
IS
fm
en
a
Conditions
Vcm ˆ VDD/2
Vcm ˆ VDD/2
Vcm ˆ VDD/2
RL ˆ 2 kO
CL ˆ 200 pF; RL ˆ 2 kO
CL ˆ 200 pF; RL ˆ 2 kO
1 kHz
100 kHz
Measured
500 mV
0.7 mV/8C
70 dB
80 dB
70 dB
90 dB
0.7 V/ms
2.5 MHz
200 mA
758
p
<35 nV/p
Hz
<16 nV/ Hz
Results shown are for nominal conditions: 3 V supply and room temperature.
currents. Fig. 11 shows the response of the sensor for an
ascorbic acid concentration of 0.02 M. Ascorbic acid has an
oxidation potential of approximately 0.6 V for platinum
electrodes. Therefore, the on-chip potentiostat was con®gured to apply a square wave with a peak-to-peak voltage of
0.8 V between the reference and working electrodes. After
charging currents subside, the sensor output current follows
the relation
C
i / K p ;
t
where C is the concentration, t the time, and K is a constant
[10]. The spikes in Fig. 11 represent charging and discharging currents. If the measurement is always taken at the
same point on the curve, the output current will be linear
with concentration. Fig. 12 plots the sensor output for
ascorbic acid concentrations from 10 6 to 0.1 M. As
expected, the response is linear.
5.4. Pressure sensor results
The initial test of the capacitive interface circuit was
performed with an 8.2 pF external reference capacitor and
known capacitances to represent the sensor; see Fig. 13. As
expected, the ®gure shows that the output voltage has a
linear response to capacitance.
Capacitive pressure sensors, however, demonstrate a nonlinear relationship between pressure and capacitance. Fig. 14
shows the results of a micromachined pressure sensor undergoing a pressure sweep from as low as 10 to around
1350 Torr. The sensor used is based on work by Chavan
and Wise at the University of Michigan [15]. This particular
non-bossed sensor experiences a change in capacitance of
only 340 fF over the pressure range.
An attempt at a linear curve ®t reveals that the sensors do
not respond linearly to pressure. A quadratic ®t, however,
produces an excellent calibration curve; the MS-8 can
support the linearization of transducer output signals using
the MAC unit.
5.5. Temperature sensor results
To test the temperature sensor, the MS-8 and test-board
were placed in an oven. The temperature was ramped slowly
from room temperature, 228C, to 808C, as measured by a
mercury thermometer. At each temperature, the chip was
Fig. 11. Output current (converted to voltage) of an amperometric sensor in a 0.02 M ascorbic acid solution. A square wave with amplitude 0.8 Vpp was
applied to the sensor. The circles denote measurement intervals. The background solution is a standard phosphate buffer with a pH of 7.6.
K.L. Kraver et al. / Sensors and Actuators A 91 (2001) 266±277
Fig. 12. Ascorbic acid calibration curve for an amperometric sensor. The response is linear over six decades of ascorbic acid concentration.
Fig. 13. Capacitive sensor readout circuit test. Fixed capacitors ranging from 1 to 16 pF were used in this test. The PGA gain is set to 1 V/V.
Fig. 14. Capacitive pressure sensor response to a pressure sweep. The plot compares the results for two PGA gain settings: 1 and 21.
275
276
K.L. Kraver et al. / Sensors and Actuators A 91 (2001) 266±277
Fig. 15. Temperature sensor response from 25 to 808C with the PGA gain set to 11 V/V.
allowed to reach equilibrium within the oven environment
before a digital voltage reading was recorded. Fig. 15 shows,
as expected, that the temperature sensor output voltage has a
linear relationship to temperature.
6. Conclusions
The MS-8 described here is the ®rst implementation of an
on-going project that coordinates efforts among analog
designers, compiler writers, and low-power digital designers
to develop a low-power, low-cost, and multipurpose sensor
interface and data acquisition system. This single-chip
microinstrument contains a programmable analog frontend capable of interfacing to a variety of sensors. The
integrated microcontroller and peripherals support digital
®ltering and compensation of sensor outputs, timing control
for sampling multiple sensors, and communication with a
host system. The usefulness of the MS-8 in embedded sensor
applications is evidenced by the initial results presented.
Acknowledgements
The authors acknowledge the ®nancial support from
National Semiconductor Corporation and the technical support from these individuals: E. Dietz, H. Verhoeven and D.
Vieira of National Semiconductor for discussions concerning the design of the opamp and review of the analog layout;
M. Lortz and M. Embacher from National Semiconductor
for support with the physical design; A. Drake from the
University of Michigan for assistance with digital testing;
and Andy DeHennis from the University of Michigan for
providing the pressure sensors.
[2] A.V Chavan, A. Mason, U. Kang, K.D. Wise, Programmable mixedvoltage sensor readout circuit and bus interface with built-in self-test,
IEEE ISSCC Digest of Technical Papers, February 1999, pp. 136±
137.
[3] H.D. Goldberg, The batch fabrication of integrated chemical sensor
arrays, Ph.D. Dissertation, The University of Michigan, Ann Arbor,
MI, 1993.
[4] A. Bindra, Ultra-low-noise data-acquisition IC tackles multiple
sensors, Electron. Design 47 (22) (1999) 59±60, 62.
[5] Analog Devices, Inc., ADuC824 data sheet, http://products.analog.com/products/info.asp?productˆADUC824.
[6] T. Cummins, E. Byrne, D. Brannick, D.A. Dempsey, An IEEE 1451
standard transducer interface chip with 12-b ADC, two 12-b DACs,
10-KB flash EEPROM, and 8-b microcontroller, IEEE J. Solid-State
Circuits 33 (12) (1998) 2112±2120.
[7] P.L. Bird, R.B. Brown, T.N. Mudge, L. Hong, B. Khailany, M.R.
Guthaus, MS-8 design specification, Technical Report, The University of Michigan, Ann Arbor, MI, 1999.
[8] R.B. Brown, P.L. Bird, M.R. Guthaus, K.L. Kraver, B. Khailany, MS8 microarchitecture, Technical Report, The University of Michigan,
Ann Arbor, MI, 1999.
[9] S.M. Sze (Ed.), Semiconductor Sensors, Wiley, New York, 1994.
[10] A.J. Bard, L.R. Faulkner, Electrochemical Methods, Wiley, New
York, 1980.
[11] P.T. Kissinger, W.R. Heineman (Eds.), Laboratory Techniques in
Electroanalytical Chemistry, Marcel Dekker, New York, 1984.
[12] B.E. Boser, B.A. Wooley, The design of sigma±delta modulation
analog-to-digital converters, IEEE J. Solid-State Circuits 23 (6)
(1998) 1298±1308.
[13] R. Hogervorst, J.H. Huijsing. Design of Low-Voltage, Low-Power
Operational Amplifier Cells, Kluwer Academic Publishers, Boston,
1996.
[14] H.J. Yoon, J.H. Shin, H. Nam, G.S. Cha, T.D. Strong, R.B. Brown,
Solid-state ion sensors with a liquid junction-free polymer membrane-based reference electrode for blood analysis, Sens. Actuators B
64 (2000) 8±14.
[15] A.V. Chavan, K.D. Wise, A batch-processed vacuum-sealed capacitive pressure sensor, in: Proceedings of the International Solid-State
Sensors and Actuators Conference (Transducers'99), Vol. 2, Chicago,
IL, 16±19 June 1997, pp. 1449±1452.
References
[1] K.D. Wise (Ed.), Scanning the issue, Proc. IEEE Integrated
Sens., Microactuators, Microsyst. (MEMS) 86 (Special Issue)
(1998) 1531±1533.
Biographies
Keith L. Kraver received the BS degree in electrical engineering from
Arizona State University in 1995. He received an MS in electrical
K.L. Kraver et al. / Sensors and Actuators A 91 (2001) 266±277
engineering (VLSI) in 1997 from the University of Michigan and is
currently a PhD candidate. His research interests include low-voltage and
low-power analog circuit design, mixed-signal design, and deep-submicron
analog circuits.
Matthew R. Guthaus received his BSE in computer engineering and MSE
in electrical engineering from the University of Michigan in 1998 and
2000, respectively. He is currently pursuing his PhD in electrical
engineering at the University of Michigan. His research interests include
low-power architecture, algorithm specific microprocessors, and computer-aided design of integrated circuits.
Timothy D. Strong received his BS degree in electrical engineering from
Michigan Technological University in 1992. After working for 3 years in
microprocessor design and verification at IBM, he began graduate studies
at the University of Michigan. He received an MS in electrical engineering
in 1997 and is currently a PhD candidate. His current research is in IC
fabrication and integrated chemical sensing via ion-selective electrodes
and cyclic voltammetry.
Peter L. Bird received the MS and PhD degrees from the University of
Michigan in 1982 and 1987. He worked as the principle hardware and
software architect for Applied Dynamics International, helping to design
and implement a development system for high speed embedded
controllers. He worked as the principle architect for the Advanced
Computer Research Institute (ACRI), a French start-up company
established to develop a high performance, heterogeneous multi-processor
system. He is currently adjunct professor of electrical engineering and
computer science at the University of Michigan where he is responsible for
the senior level course in compiler design and implementation. His
research interests include language design and implementation, application
of parsing techniques to software engineering problems, and computer
architectures for high performance and embedded systems. He holds a US
patent for a high performance memory system.
277
Geun Sig Cha received his BS degree from Korea University in 1977 and
his PhD in analytical chemistry from the University of Michigan in 1989.
He continued his research at the University of Michigan as a joint research
fellow in the Department of Chemistry and Electrical Engineering and
Computer Science until January 1991. He is now an associate professor of
analytical chemistry in the Department of Chemistry at Kwangwoon
University. His current research interests involve the development of
polymer membrane-based ion-, bio- and gas-sensors and the design of
enzyme-linked competitive and non-competitive binding assay methods.
Wolfgang HoÈld was born on 30 May 1965 in Guenburg, Germany. He
received the Dipl. Ing. in electrical engineering (focus on communication
electronics) in 1989 from the Technical University, Augsburg, Germany. In
1989, he joined National Semiconductor, Germany, as a product engineer
for 4- and 8-bit microcontrollers. From 1990 to 1994, he held positions in
product engineering and design engineering for microcontrollers at
National Semiconductor, Santa Clara, CA. In 1994, he established a
design center for microcontrollers at the National Semiconductor
European Headquarters, Germany. The design center focused on 8- and
16-bit microcontroller cores and communication modules. His current
interests and activities include developing best design practices for HDL
coding and design reuse standards. He is presently responsible for the 16bit Compact-RISCTM CPU, including product development.
Richard B. Brown received his BS and MS degrees in electrical
engineering from Brigham Young University in 1976. He then worked in
computer design and instrumentation until returning to school in 1981. He
received an electrical engineering PhD at the University of Utah in 1985.
His dissertation was on integrated solid-state chemical sensors, which
included development of a custom MOS fabrication process and
integration of multiple transducers with analog and digital circuitry. In
September 1985, he joined the Department of Electrical Engineering and
Computer Science, University of Michigan, where he serves as associate
chair. His areas of current research are SOI circuits, mixed-signal
integrated circuits, and solid-state chemical sensors.